SDN

OpenFlow-enabled
Transport SDN
ONF Solution Brief
May 27, 2014
O N F S O LU T I O N B R I E F
O p e n F l o w - e n a b l e d Tr a n s p o r t S D N
Table of Contents
2
Executive Summary
SDN Overview
Trends and Challenges
Enhancing Transport Networks with OpenFlow
Use Case: Bandwidth on Demand
10 Use Case: Private Optical Networks

11 Use Case: Optical Virtual Private Networks
13 Use Case: Multi-vendor IP/MPLS Plus Transport Optimization
15 Key Benefits of Transport SDN
15 Conclusion
16 Contributors
16 References
Executive Summary
Wide area network (WAN) traffic continues to double every 18 to 24 months, with
individual service flows now exceeding 1Gbps and inter-site aggregates reaching
tens to hundreds of gigabits per second. Simultaneously, mobility networks and
cloud-based services are causing traffic patterns to become ever more dynamic and
unpredictable. In the face of these realities, todays static and manually configured
transport networks operating in their own silos are woefully inadequate. Transport
networks need to become more flexible and dynamic to support end-user demands.
OpenFlow-based Software Defined Networking (SDN) initially enabled open,
application-driven, programmatic control of packet forwarding among a set of packet
switches. However, not all services are packet-based, and very few WANs are
constructed with packet switches directly connected by fibers as may be the case in
data centers. Most deployed WANs employ SONET/SDH or OTN sub-wavelength and
wavelength circuit-switching elementstechnologies that are not currently supported
by OpenFlow. These elements provide reliable, efficient transport services directly to
end users or to client packet nodes (e.g., IP/MPLS routers) that, in turn, provide packet
services to end users.
Recently, packet transport technologies such as MPLS-TP have been incorporated
with optical transport, resulting in more cost-effective packet-optical transport systems
(P-OTS). Based on Heavy Readings March 2013 Metro P-OTS survey1, revenue of
such P-OTS elements in metro networks already exceeds that of SONET/SDH nodes
and is projected to outpace that of pure DWDM systems by 2015.
Open Networking Foundation. All rights reserved.
2 of 16
This solution brief addresses how OpenFlow can enable open, application-driven,
programmatic control of optical and packet-optical transport elements. It also
offers several illustrative use cases, including bandwidth on demand, private optical
networks, optical VPN services, and IP/MPLS plus Transport optimization.
SDN Overview
Software Defined Networking is a new architecture that has been designed to enable
more agile and cost-effective networks. The Open Networking Foundation (ONF) is
taking the lead in SDN standardization, and has defined an SDN architecture model as
depicted in Figure 1.
FIGURE 1
ONF/SDN architecture
APPLICATION LAYER
Business Applications
API
CONTROL LAYER
Network
Services
API
API
Network Services
INFRASTRUCTURE
LAYER
The ONF/SDN architecture consists of three distinct layers that are accessible through
open APIs:
The Application Layer consists of the end-user business applications that
consume the SDN communications services. The boundary between the
Application Layer and the Control Layer is traversed by the northbound API.
The Control Layer provides the logically centralized control functionality that
supervises the network forwarding behavior through an open interface.
The Infrastructure Layer consists of the network elements (NE) and devices that
provide packet switching and forwarding.
3 of 16
According to this model, an SDN architecture is characterized by three key attributes:

Logically centralized intelligence. In the ONF SDN architecture, network
control is distributed from forwarding using a standardized southbound interface:
OpenFlow. By centralizing network intelligence, decision-making is facilitated
based on a global (or domain) view of the network, as opposed to todays
networks, which are built on an autonomous system view where nodes are
unaware of the overall state of the network.
Programmability. SDN networks are inherently controlled by software
functionality, which may be provided by vendors or the network operators
themselves. Such programmability enables network configuration to be automated,
influenced by rapid adoption of the cloud. By providing open APIs for applications
to interact with the network, SDN networks can achieve unprecedented innovation
and differentiation.
Abstraction. In an SDN network, the business applications that consume SDN
services are abstracted from the underlying network technologies. Network
devices are also abstracted from the SDN Control Layer to ensure portability and
future-proofing of investments in network services, the network software resident
in the Control Layer.
Trends and Challenges

Transport networks are under pressure. Demand for bandwidth continues to grow rapidly
with no apparent end in sight. Traffic patterns are seismically shifting due to the adoption
of cloud services, the dramatic increase in video usage, and the emergence of mega-sized
data centers.
At the same time, device mobility and an Internet of things has changed where and how
bandwidth is being consumed. Self-serve, on-demand infrastructure, and applications
based on virtualized compute and storage, have changed customer expectations of the
WAN. Utilization is dynamic, and the network needs to support that dynamism. Traditional
L2/L3 VPN services provide the necessary flexibility, but not the level of determinism or
control required. High peak-to-average and/or transient bandwidth demands between
certain locations require transport services that can be turned up, modified, and torn down
in near real time.
4 of 16
Current transport networks cannot effectively address these pressure points, as they are
generally static and operated separately from the client layers and applications they serve.
Traditional transport services can take weeks or months to turn up and must be contracted
for periods of months or yearsa reflection of the planning and provisioning effort currently
endured. Because of the lengthy process to turn up new services, many customers cancel
orders before they become operational.
To support the highly dynamic environment mentioned above, connectivity services must be
turned up in minutes or seconds and be modifiable by client users or software applications
without operator intervention. Orchestration is required to manage connectivity services
over a network covering potentially multiple network domains, through multiple layers of
networking technology, and across multiple vendors equipment.
Enhancing Transport Networks with OpenFlow

Transport SDN, which is being defined under the umbrella of ONFs Optical Transport
Working Group (OTWG),2 provides the architecture and mechanisms to address these
trends and challenges. In March 2013, ONF chartered the OTWG to extend the OpenFlow
standard to support transport networks . The OTWG has liaised with peer standards
bodies such as the OIF, ITU-T, BBF, and TMF to reach alignment on the target use cases,
architectural proposals, and information model.
Through automation and logically centralized intelligence, OpenFlow-based Transport
SDN simplifies operators complex, rigid, and multi-vendor environments and enables the
introduction of new transport network services to better monetize the network.
Until recently, the OpenFlow standard focused on the packet-oriented Layers 2 and 3.
Transport SDN extends OpenFlow to support Layer 0 (photonic) and Layer 1 (SONET/SDH,
OTN) networks, allowing the same support for logically centralized control and independent
software and hardware development. Fundamentally, extensions are added to OpenFlow
to program switch ports and fabrics that operate on fibers, wavelengths, and timeslots as
well as packet headers, while retaining the same simple model of match and action tables
across these multiple layers.
To address the specific requirements for transport networks, OpenFlow is also being
extended to support protection, performance monitoring, and other critical operations,
administration, and maintenance (OAM) capabilities. Such performance monitoring
information (e.g., bit error rate) could be used for consequential dynamic service control. See
Figure 2 for a summary of Transport SDN extensions.
5 of 16
EXTENTIONS
L0/L1-capable port descriptions

Including neighbor discovery / ID
FIGURE 2
Target multi-layer OpenFlow
logical switch model
Match table capable of matching/

changing L0/L1 "fields", adding MEPs
Match & Group tables supporting
NE-based protection and OAM
Dynamic multilayer control of
switching and adaptations, e.g.
Ethernet > ODUk > Optical Channel
Ethernet > ODUj > ODUk >
Optical Channel
Ethernet > MPLS LSP > ODUk >
Optical Channel
The application of these L0/1/2 OpenFlow extensions, as elucidated in the use case sections
below, includes:
Packet-optical integration (POI) to support logically centralized control of multi-layer
and/or multi-technology networks, resulting in optimized network utilization.
Open transport network control programmability to spur greater
network innovation.
Real-time application-driven provisioning of the transport network.
An important architectural innovation is the introduction of transport network
virtualization, whereby the operators physical transport network can be partitioned
into multiple virtual networks. Each virtual transport network is allocated certain virtual
network elements (V-NEs) and interconnecting virtual links (VLs) that are under the
control of the respective clients through their client controllers.
The ONF Transport SDN model defines two interface points where OpenFlow can
be applied:
The CDPI (control data plane interface), originally defined in an ONF white paper 3,
where there is a direct interaction between the network operators SDN controller
and the transport network elements.
The CVNI (control virtual network interface), where the client controllers use
OpenFlow to interact with the providers controller that offers a virtualized view of
each clients slice of network resources.
6 of 16
A client controller can belong either to an internal customer (e.g., the operators own content
distribution network), or to an external customer (such as an enterprise or another operator).
The scope of the clients visibility into their resources is still under definition, but may vary
from completely abstracted to completely specific depending on the policy set by the
provider, which is generally a function of the trust relationship with the client.
FIGURE 3
Hierarchical control
architecture
The hierarchical controller structure illustrated in Figure 3 enables control to be explicitly

partitioned across the physical network resources. Hierarchy further provides a means for
coordinating multiple domains in an operators network for greater scalability and for
integration across different layers or technologies. The use of CVNI also may ease the
migration process or coexistence between legacy transport networks and OpenFlow-based
transport networks. The control plane under the CVNI could be anything, including
distributed control planes such as GMPLS. A mediation application to translate to traditional
transport management interfaces (e.g., TL1 or CORBA) can be introduced to allow legacy
transport domains to operate in an SDN environment.
7 of 16
In addition to greater network resource and operational efficiencies, Transport SDN allows
network programmability for the rapid deployment of new transport services, enabling
operators to better differentiate and monetize their infrastructures. Examples of such
offerings are tiered recovery, customized path selection, and various applications that will
make the transport network aware of packet (client) layers and their dynamic requirements in
terms of bandwidth, latency, geography, and time span.
Use Case: Bandwidth on Demand

The increases in big data, cloud networking, and ad hoc inter-enterprise collaboration result
in inter-site bandwidth peaks that can exceed the mean by 10x, 20x, or moreand can last
anywhere from tens of minutes to several hours or even longer. Contracting for a private line
or committed information rate (CIR) sized to accommodate such peaks is at best wasteful
and for some, prohibitively expensive. Research4 shows that approximately 35% of key
industry verticals such as finance, manufacturing, media and entertainment, and others are
very interested in dynamic services that address this conundrum.
OpenFlow-based Transport SDN enables carriers to offer bandwidth on demand (BWoD)
services that allow customers to dynamically establish and subsequently resize connectivity
between their sites as necessaryand to pay only for the network resources that they
actually use.
These services require access from each customer site to a dedicated port at the edge
of the carrier switched network (either at customer premises or backhauled over a fixed
access segment). In either case, the UNI is typically Ethernet (e.g., 10/100GbE) or OTN (e.g.,
OTU2/3/4), though others are possible. Forwarding from the UNI may be physical portbased or logical sub-portbased, in which case customer traffic is mapped to its respective
BWoD switched virtual leased line (SVLL) by matching ingress VLAN ID, MPLS label,
SONET/SDH AU-4/STS-1 time slot, OTN OPUk tributary slot, or wavelength frequency slot.
Inter-site switching of the SVLL will be OTN ODU or wavelength-based.
8 of 16
FIGURE 4
BWoD service illustrating
port-based (above)
and sub-port based
(below) forwarding
On-demand connections are established by the customer interacting with a bandwidth

broker/scheduler application hosted on an SDN controller. Requested connection
parameters are communicated through a portal or over a northbound API (NBI). The
northbound API allows operations users, customer clients, and applications to request
SDN communications services directly from the Control Layer. Connection parameters may
include source and destination UNIs, bandwidth, maximum latency, protection tier (e.g.,
unprotected, 1+1, 1:1, or mesh restorable), and sub-port mapping identifiers (e.g., VLAN ID),
if applicable. In addition, the connection request may be for immediate instantiation or a
specified scheduled time in the future, and for an indefinite or specified duration.
The broker/scheduler application first validates the request against the customers
contracted policy (e.g., endpoints or maximum aggregate bandwidth) and then invokes
the path computation element (PCE) to determine whether a viable path exists. If so, it
proceeds to instantiate the connectioneither immediately or when scheduledvia the
SDN controller. The SDN controller again uses the PCE to identify the best path and issues
OpenFlow commands to the transport NEs to establish the bi-directional connection,
including the mapping from client port to connection origination point (e.g., ODUj/k), the
hop-by-hop ODU and/or wavelength cross-connects, and the connection termination point
9 of 16
to client port de-mapping. Multiple sub-port connections can be supported by the UNI
simultaneously, and the resources of released or decremented connections are immediately
available for other connections.
Transport SDNenabled bandwidth on demand provides benefits to end customers and
network operators alike. End customers can receive high bit-rate, deterministic capacity on
a temporary basis when and where they need it, without the cost burden of over-provisioned
private lines with long-term contracts. Priced intelligently, the network operator increases
overall revenues by increasing their addressable market to those who cant afford such
contracts. At the same time, the operator increases return on assets by selling capacity
several times over per month and for higher value per time unit than a flat monthly fee yields.
Use Case: Private Optical Networks

There have always been enterprises for which the network is such a strategic part of their
business that they have chosen to build and manage their own private optical networks.
Such enterprises directly procure networking equipment and connect their locations with
leased dark fibers.
An enterprise private optical network is often a microcosm of the public network. Some
enterprises connect large offices and data centers for cloud-based virtual compute and
storage. Some enterprises support mobile users. These enterprises experience high peakto-average inter-site bandwidth demands, and require rapid capacity reallocation driven by
applications such as cloud orchestration systems. Consequently, enterprises running private
optical networks may also benefit from Transport SDN.
SDNs logically centralized control provides a network-wide perspective to optimize the
use of all resources through the network. SDN also enables service automation to simplify
and accelerate provisioning and reduce potential for human errors. With Transport SDN,
enterprises can switch wavelengths to allocate bandwidth among sites as needed by their
applications (see Figure 5). For example, they can:
Redirect wavelengths through ROADMs.
Retune transceivers on optical transponders/muxponders or colored interfaces on
OpenFlow-enabled switches/routers to a new wavelength.
Activate new connections and deactivate old connections.
10 of 16
FIGURE 5
Private optical network with
reconfigured wavelengths
Until dynamic optical services become widely available from local providers, Transport
SDN enables enterprise IT departments to function as their own service providers.
Such in-house capability enables them to manage connectivity and allocate
bandwidth between sites in minutes rather than waiting weeks or months for todays
typical local provider to fulfill their fixed-line change order. Instead of ordering fixed
bandwidth services sized for the peak, an enterprise managing a private optical
network has the flexibility to manage their network and orchestrate capacity shifts
among their locations where and when they are needed.
Use Case: Optical Virtual Private Networks

While many enterprises are interested in the benefits of private optical networks, not all
can afford the dedicated equipment and leased fibers of a bespoke network operated
by a service provider, or have the in-house expertise to manage a fully private optical
network on their own. Such enterprises are prime candidates for optical virtual private
network (O-VPN) services. Extending the concept of VPN to the optical layers is not
an entirely new concept. There are commercial solutions available today; however,
most are based on sophisticated, vendor-specific management systems and only
allow end users to provision wavelength or sub-wavelength connections across their
VPN resources through a portal.
The key challenge for IP/MPLS VPN services is to provide deterministic performance.
Optical network-based services are circuit-based and inherently deterministic.
However, there are two challenges with O-VPNs: providing independent control and
management across the end-to-end service, and provisioning the O-VPN services
across various layers and vendor network elements. Both of these challenges are
addressed by Transport SDN.
11 of 16
Using Transport SDN, O-VPN service is made possible through network virtualization
(described above) that allows a network operator to create virtual slices of the network
that deliver dedicated capacity and provide the client with self-managed control of
their end-to-end virtual networks. With O-VPN, the client can allocate capacity as if it
were their own private network.
FIGURE 6
Common optical network
supporting multiple O-VPNs
OpenFlow-enabled Transport SDN allows network operators to expose the underlying

multi-layer, multi-vendor network as a virtual topology. The visibility and performance
information (e.g., latency) of this topology over which connections can be created can be
further controlled for full or partial view. For example, Figure 6 shows an optical network
that supports multiple O-VPNs with different levels of virtual network granularity.
The connection creation can be achieved either through a client portal or from a
clients controller over the CVNI by specifying connection details, including end points,
size, explicit path, resiliency, protection path, and more. This results in a dynamic VPN
that allows clients to develop their own algorithms to manage connection paths in
support of their unique requirements.
12 of 16
For example, a distributed, multidata-center cloud operator could use Transport SDN
to orchestrate capacity turn up and tear down between specific data centers to support
bulk VM migration or the synchronization of large databases. Based on the ebb and flow
of bandwidth demands, portions of the same O-VPN capacity could be reallocated to a
different application or a different pair of data center locations as needed.
The key benefit that Transport SDN provides for O-VPN end users is access to a
deterministic network at a much lower cost. For network operators, the key benefits
are reduced costs due to sharing network elements, fiber, and wavelengths across
multiple internal and external customers, and higher profits based on the introduction
of new value-added services with higher revenue and better margins than simply
leasing dark fibers.
Use Case: Multi-vendor IP/MPLS Plus Transport Optimization

Presently, IP/MPLS services deliver most application-based, content-rich services.
Because it is inefficient to interconnect routers with dedicated fibers, operators have
deployed Layer 0/1 transport networks to deliver and manage capacity. With the traffic
growth and dynamic patterns described in this solution brief, transport networks must
become more dynamic. This use case describes coordinating multi-layer IP/MPLS and
transport provisioning and restoration to address shifting aggregate IP flows.
In network parlance, the IP/MPLS routers are clients and the transport nodes are
servers. Presently, most IP/MPLS topologies assume a static transport layer. Physical
router topologies and transport connections are sized based on 6- to 12-month traffic
forecasts, worst-case source-to-destination peaks, and allowances for extra capacity
to recover from link or IP/MPLS node or port failures. The IP/MPLS traffic is 1+1
protected, and the transport network provides two diverse optical paths to the routed
network. The result is IP/MPLS networks that are no more than 40% efficient.
This is largely a byproduct of the fact that current IP/MPLS plus transport networks
typically have disjoint provisioning processes and network management systems. IP/
MPLS services and transport services are independently provisioned using separate
workflows. Path computation and traffic engineering are employed at each layer of the
solution separately. Control plane information for each layer is also held separately in
the IP/MPLS routers and the transport nodes.
Transport SDN provides the mechanisms for a dynamic transport layer and
defines an architecture to enable the IP/MPLS and transport layers to be controlled
synergistically. Typical adaptation between domains includes either IP/MPLS
pseudowire or MPLS-TP tunnel to GFP-F to ODU to DWDM physical port, and could
be performed on a router with colored interfaces or a packet-optical transport node.
13 of 16
FIGURE 7
Multi-layer IP/MPLS plus
transport control
The solution may use a single multi-layer controller interfacing directly to both IP/MPLS
and transport elements, or separate domain controllers and a multi-layer orchestrator/
controller, as illustrated in Figure 7. In this approach, each domain controller provides
detailed information about topology, latency, provisioned services, and performance
to the multi-layer controller for path computation and restoration management.
When bandwidth shortfalls are anticipated, a failure occurs, or more efficient routes
are possible, the multi-layer controller can orchestrate changes to the IP/MPLS and
transport layers through the respective domain controllers, by allocating or reallocating
router ports and transport capacity, rerouting transport connections, or creating
express routes.
This multi-layer IP/MPLS + transport optimization achieves many network benefits,
including:
CapEx reduction, by reducing the need for over-provisioning of the network to
support demand shifts and protection/restoration.
Increased service availability (e.g., coordinated protection and restoration) and
service quality (e.g., latency-optimized multi-layer provisioning).
14 of 16
Multi-layer control and optimization across separate IP/MPLS and transport vendors.
OpEx reduction and simplification through automation to reduce manual processes
and associated configuration errors.
Increased revenues, by leveraging network intelligence to monetize the network
based on a broad list of programmable path and service level parameters, such as
end-to-end latency of an IP/MPLS service.
Key Benefits of Transport SDN

OpenFlow-enabled Transport SDN provides several benefits to optical and packetoptical networking, including:
Speed and automation. Programmatic APIs and logically centralized control
streamline operations compared to manual order entry and NMS-based
provisioning.
Operator innovation. A standardized information model and open northbound
APIs allow network operators to develop their own service models and constraintbased connection routing algorithms to differentiate their services.
Efficiency. Using OpenFlow for conjoint control of packet, circuit, and photonic
transport resources enables multi-layer path computation and optimization to
intelligently aggregate and express traffic to minimize stranded capacity.
Multi-vendor networking. Standard OpenFlow enables common connection
provisioning commands across any compliant NE, allowing the choice of different
vendors for different layers and/or geographical or operational domains.
Conclusion
In this era of intense competition, high bandwidth demand growth, and unpredictably
shifting traffic patterns, operators need their transport networks to become
dynamically programmable in order to offer new services and to match capacity
to traffic demand without over-provisioning network resources. Transport SDN,
as defined by the ONF, provides the architecture and mechanisms to enable
programmatic and dynamic control driven by the revenue-generating services that
consume network capacity. This linkage, governed by policy management, will
enable network operators to transform their transport networks to simultaneously
reduce capital and operational expenses while increasing revenue opportunities and
operational agility.
15 of 16
Contributors
Mitch Auster, Editor
Sivaram Balakrishnan
Mike McBride
Greg Nehib
Lyndon Ong
Meral Shirazipour
Vishnu Shukla
Abel Tong
Maarten Vissers
References
1. Metro Packet-Optical Transport 2.0: A Heavy Reading Survey Analysis
http://www.heavyreading.com/details.asp?sku_id=2988&skuitem_itemid=1477
2. Charter: Optical Transport Working Group
https://www.opennetworking.org/images/stories/downloads/working-groups/charter-opticaltransport.pdf
3. Software-Defined Networking: The New Norm for Networks
https://www.opennetworking.org/images/stories/downloads/sdn-resources/white-papers/wp-sdnnewnorm.pdf
4. The Power of Software in Cienas OPn Network Architecture
http://www.ciena.com/resources/white-papers/The-Power-of-Software-in-Cienas-OPn-NetworkArchitecture.html
Open Networking Foundation / www.opennetworking.org

The Open Networking Foundation is a nonprofit organization founded in 2011, whose goal is to accelerate the adoption of open SDN.
ONFemphasizes the interests ofendu sers throughout the Data Center, Enterprise, and Carrier network environments.
Open Networking Foundation, the ONF symbol, and OpenFlow are registered trademarks of the Open Networking Foundation, in the United States
and/or in other countries. All other brands, products, or service names are or may be trademarks or service marks of, and are used to identify,
products or services of their respective owners.
16 of 16

SDN

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

SDN

Hochgeladen von

Copyright:

Verfügbare Formate

OpenFlow-enabled

Trends and Challenges

Enhancing Transport Networks with OpenFlow

Use Case: Bandwidth on Demand

10 Use Case: Private Optical Networks

Open Networking Foundation. All rights reserved.

Open Networking Foundation. All rights reserved.

According to this model, an SDN architecture is characterized by three key attributes:

Trends and Challenges

Open Networking Foundation. All rights reserved.

Enhancing Transport Networks with OpenFlow

Open Networking Foundation. All rights reserved.

L0/L1-capable port descriptions

Match table capable of matching/

Open Networking Foundation. All rights reserved.

The hierarchical controller structure illustrated in Figure 3 enables control to be explicitly

Open Networking Foundation. All rights reserved.

Use Case: Bandwidth on Demand

Open Networking Foundation. All rights reserved.

On-demand connections are established by the customer interacting with a bandwidth

Open Networking Foundation. All rights reserved.

Use Case: Private Optical Networks

Open Networking Foundation. All rights reserved.

Use Case: Optical Virtual Private Networks

Open Networking Foundation. All rights reserved.

OpenFlow-enabled Transport SDN allows network operators to expose the underlying

Open Networking Foundation. All rights reserved.

Use Case: Multi-vendor IP/MPLS Plus Transport Optimization

Open Networking Foundation. All rights reserved.

Open Networking Foundation. All rights reserved.

Key Benefits of Transport SDN

Open Networking Foundation. All rights reserved.

Open Networking Foundation / www.opennetworking.org

Das könnte Ihnen auch gefallen